Category: networking

My Home Networking Setup

I’ve had a few asks from various people about my networking setup at home so I thought I’d throw together a blog post detailing it.

I wanted something that was going to be stable, performant, and reliable so I bit the bullet and decided to go with equipment that typically isn’t used in most households. No Linksys or Netgear consumer grade equipment here!

My current setup consists:

1 x Ubiquiti Edge Router ERPro-8
1 x Ubiquiti Edge Switch ES-24-250W
4 x Ubiquiti Unifi 802.11ac Dual-Radio PRO Access Point (UAP-AC-PRO-US)

My ER-8 Pro (datasheet) is connected to AT&T Fiber for a lightning fast 1GB up and down. The router has six ports which is probably overkill for a home set up but this version had a faster MIPS processor and was rack mountable. There’s also a non pro version but it has a slightly slower processor. Eventually I may add a secondary failover provider to ensure the internet is always up and keep the family happy at all times. The router also runs Debian which lets you ssh into the router and use the CLI to configure or hack around on the router.

The 24 Port PoE Switch is a relatively cheap switch that provides PoE (Power over Ethernet). All of the AC-Pro Wifi access points are powered with PoE from the switch and distributed around the house for maximum signal. This is great because installation of devices just requires an ethernet line instead of the additional power. An added benefit is that all of the equipment can be battery backed up from one central location. My home is hard wired for ethernet so all wiring aggregates in one location to the switches. Wifi is usually nice to have for roaming devices, but for dedicated devices that usually require higher bandwidth, like gaming consoles, Apple TVs, and the TVs themselves, it’s nice to have those hardwired into the network.

Overall I’ve been really happy with the setup and haven’t really had to mess with it much other than the occasional firmware updates to keep the devices up to date.

Setting up an OpenStack Cloud using Ansible

I use Ansible and OpenStack quite a bit on a daily basis, so I wanted to check out the work the community has done with the openstack-ansible project and get an OpenStack environment set up in my lab. I encourage you to read through the documentation as it is really detailed. Let’s do this!

img

My Lab Environment

My setup consists of:

4 x Dell PowerEdge R720s with 128GB of RAM
Quad 10G Intel NICs
Cisco Nexus 3k switches

I set aside one of the nodes for deployment and the other three were going to be used as targets. openstack-ansible currently supports Ubuntu 14.04 LTS (Trusty) so the first order of business was to install the OS to the servers. Future support for 16.04 LTS (xenial) and CentOS 7 may be coming down at some point as well.

Setting up Networking

Once the OS was installed, the first thing to do was to set up the initial networking config in /etc/network/interfaces. For my setup, I’ll be assigning networks to vlans for my setup.

Add some initial packages on the target host and enable some modules:

apt-get install bridge-utils debootstrap ifenslave ifenslave-2.6 \
lsof lvm2 ntp ntpdate openssh-server sudo tcpdump vlan
echo 'bonding' >> /etc/modules
echo '8021q' >> /etc/modules

Drop your interfaces file onto all of your hosts you’ll be deploying and reboot them to apply the changes so that they set up all of the bridges for your containers and instances. In my example this configuration sets up dual bonds, VLANs, and bridges that OpenStack Ansible will plug everything into.

Initial Bootstrap

You’ll want to use one server as the deployment host so log into that server, check out openstack-ansible, and run the initial Ansible bootstrap:

git clone https://github.com/openstack/openstack-ansible.git /opt/openstack-ansible
cd /opt/openstack-ansible
git checkout stable/mitaka
scripts/bootstrap-ansible.sh

The bootstrap-ansible.sh script will generate keys so make sure to copy the contents of the public key file on the deployment host to the /root/.ssh/authorized_keys file on each target host.

Copy the example openstack_deploy directory to /etc/:

cp -R /opt/openstack-ansible/etc/openstack_deploy /etc/openstack_deploy
cp /etc/openstack_deploy/openstack_user_config.yml.example /etc/openstack_deploy/openstack_user_config.yml

Modify the openstack_user_config.yml for the settings you want. You’ll need to specify which servers you want each role to do. The openstack_user_config.yml is pretty well commented and provides lots of docs to get started.

My config:

If you have enough memory and CPU on the hosts, you can also reuse the infrastructure nodes as compute_nodes to avoid having to set up dedicated nodes for compute.

Credentials

cd /opt/openstack-ansible/scripts
python pw-token-gen.py --file /etc/openstack_deploy/user_secrets.yml

User Variables

We’ll start with just the basics for now to get operational. Make sure to enable at least a few options in the /etc/openstack_deploy/user_variables.yml otherwise it will have a hard time assembling the variables (these haven’t made it into the mitaka stable branch yet):

## Debug and Verbose options.
debug: false
verbose: false

Run the playbooks

cd /opt/openstack-ansible/playbooks
openstack-ansible setup-hosts.yml
openstack-ansible haproxy-install.yml
openstack-ansible setup-infrastructure.yml
openstack-ansible setup-openstack.yml

If there are no errors, then the initial cluster should be setup. The playbooks are all idempotent so you can rerun them at anytime.

Using the Cluster

Once these playbooks complete, you should have a functional OpenStack Cluster. To get started, you can log into Horizon with either the external vip IP you set up in openstack_user_config.yml or by hitting the server directly.

You’ll use the user name “admin” and the password will be in your /etc/openstack_deploy/user_secrets.yml file that you generated earlier:

grep keystone_auth_admin_password /etc/openstack_deploy/user_secrets.yml
keystone_auth_admin_password: 4lkwtwtpmasldfqsdf

Each target node will have a utility container that you can ssh into to grab the openstack client credentials or run the client from the container. You can find it by doing an:

root@osa-node1:~# lxc-ls | grep -i util
node1_utility_container-860a6cd9
root@osa-node1:~# ssh root@node1_utility_container-860a6cd9
Welcome to Ubuntu 14.04.4 LTS (GNU/Linux 3.13.0-85-generic x86_64)
root@node1-utility-container-860a6cd9:~# openstack server list
+--------------------------------------+-------------+--------+----------------------+
| ID                                   | Name        | Status | Networks             |
+--------------------------------------+-------------+--------+----------------------+
| 1b7f1a7f-db87-47fe-a884-c66875ceed00 | my-instance | ACTIVE | Public=192.168.20.165|
+--------------------------------------+-------------+--------+----------------------+

Create and Setup Your Network

In Horizon under the System tab, select networks and then “+Create network”. The main thing to note is depending on the network you are setting up, make sure to specify that type in the Physical Network box as well. In my case, I set up a vlan network, so I made sure to set:

Name: Public
Project: admin
Provider Network Type: VLAN
Physical Network: vlan
Admin State: UP

Once the network is created, click on the Network Name and click “+Create Subnet”. Add your:

Subnet Name: VLAN_854
Network Address: 10.127.95.0/24
Gateway IP: 10.127.95.1
Allocation Pools: <Start Address>,<End Address>
DNS Name Servers: <DNS Servers>

Add Images to Glance

You’ll need to add some images to get up and running. You can find a list of supported images that include Cloud-Init here.

Name: Image Name
Image Source: Image Location
Image Location: Enter in URL of Cloud Image
Format: QCOW2, RAW, or whatever the image format may be
Architecture: x86_64 or whatever hardware you might be using
Public: Checked

Security Groups

By default security groups are enabled, so you’ll want to enable some ingress rules like SSH and ICMP by default so you can connect to your instance.

Start an Instance

Under the instances tab, click “Launch Instance”. Fill in your desired options, including boot from image, add any keypairs you might want, and make sure to select the Security Group you set up previously. You’ll also want to make sure you are plugged into the right network as well. Once all of those things are set up, you should be able to launch the instance and attempt to connect to it.

Things to Note

Cluster Recovery

The target hosts are in a DB cluster so if you need to reboot them, make sure to stagger them, so that the cluster doesn’t fail. If you find the DB is not coming up, you can run the galera-bootstrap playbook which should bring the cluster back up (docs):

openstack-ansible galera-install.yml --tags galera-bootstrap

If you run into any issues running through this, please let me know in the comments or ping me in #openstack-ansible on Freenode as antonym.

Configuring bonding with systemd

With release of systemd version 216, a number of bugs were corrected and configurations options added related to bonding support in systemd-networkd.

Here’s a basic example on setting up bonding in systemd:

/etc/systemd/network/bond0.network:

[Match]
Name=enp1*

[Network]
Bond=bond0

You can also specify the interfaces like enp1s0f[01] if you need to be more specific.

/etc/systemd/network/bond0.netdev:

[NetDev]
Name=bond0
Kind=bond

[Bond]
Mode=802.3ad
LACPTransmitRate=fast
MIIMonitorSec=1s
UpDelaySec=2s
DownDelaySec=8s

/etc/systemd/network/Management.network:

[Match]
Name=bond0

[Network]
Address=192.168.20.20/24
Gateway=192.168.20.1
DNS=8.8.8.8

Enable and start systemd-networkd:

systemctl enable systemd-networkd
systemctl start systemd-networkd

For more options, make sure to check out the systemd.network and systemd.netdev man pages.

The Dreaded Flipping of NICs

I recently had a problem with NICs flipping around after removing all traces of MAC address rules from the server. I did this because I wanted the flexibility to be able to swap machines around at any point in time and not have to worry about tracking the MAC addresses on all of the devices. The gear was identical in specifications and after doing some research, I ran across a solution that has worked really well so far. It involves creating udev rules that don’t contain any MAC addresses but that instead check the vendor id and bus location of the device. By knowing these items, you can guarantee you’ll always have the correct ethernet device assigned to the correct physical network and you can make the rules a lot more generic in nature. As an example, first you’ll want to identify the devices (example is from an HP ProLiant DL385):

1
2
3
4
lspci | grep -i eth
04:00.0 Ethernet controller: Broadcom Corporation NetXtreme II BCM5708 Gigabit Ethernet (rev 12)
42:00.0 Ethernet controller: Broadcom Corporation
NetXtreme II BCM5708 Gigabit Ethernet (rev 12)

We’ll take the first line for this example and break it down. The first group of numbers is the bus number (04), device number (00), and function
number (0). From here we should be able to generate our udev rules file. Create /etc/udev/rules.d/70-persistent-net.rules and enter in the following or whatever your setup looks like:

1
2
SUBSYSTEM=="net",ACTION=="add",BUS=="pci",KERNEL=="eth*",ID=="0000:04:00.0",NAME="eth0"
SUBSYSTEM=="net",ACTION=="add",BUS=="pci",KERNEL=="eth*",ID=="0000:42:00.0",NAME="eth1"

Once that’s in place, you should be able to reboot and not have to worry about the NICs flipping around. If you’re curious, you can also view more device information by looking at /sys:

1
ls -la /sys/bus/pci/devices/0000:04:00.0

I’ve had success with this in Citrix XenServer (dom0 is based on CentOS) and Debian.